Drowning in Data

A panel of buy side market participants at Profit & Loss Scandinavia in Stockholm shared their insights into their FX activities and discussed how data has come a long way in terms of helping them measure best execution; however, analysing that data isn’t as easy as they’d like.

Is the availability of data and technology changing how the industry assesses execution quality? That’s the question addressed by Marcus Samuelsson, portfolio manager at Ericsson; Andreas Wollheim, head of trading and treasury at SEB Investment Management; and James Koutoulas, CEO at Typhon Capital Management, speaking at Profit & Loss Scandinavia in September.

While data collection and dissemination improves, buy side managers are reporting difficulties in terms of analysing what’s available. “There’s a lot of data that wasn’t around a couple of years ago, yet we don’t really have the resources to analyse everything, so the data is difficult to assess – we’re drowning in data,” said Samuelsson.

“We have been doing transaction analysis in equities for a long time, and have started doing it in FX recently. The problem is, if you don’t know what the reference price is when you do the trade, it’s difficult to measure,” added Wollheim. “The quality of data is going to make a big difference as it improves and TCA providers become more professional.”

So how can investors develop more dynamic processes to use the data and get better TCA? The determinant lies in what is measured, noted Wollheim. The instinct, he said, is to measure the spread, but for SEB IM, when a trade is executed during the day is probably more important.

While some asset managers consider FX a throwaway transaction, speculative traders constantly look at ways to minimise slippage. Therefore, execution quality can mean many things to many people.

“Trading FX is definitely important for a corporate like Ericsson doing business globally in a lot of different markets simultaneously,” said Samuelsson, “but at the same time, resources are not finite. So, for us, we may not talk about every single transaction, but we try to focus on process and having an execution structure that allows us to quickly make decisions in different markets at different times. Flexibility is key.”

Wollheim added that the last seven to eight years has seen a shift from FX really being a residual, to having more of an appreciation about the impact FX can have on overall performance in equity or fixed income portfolios. “We are trying to motivate people by showing them the data. For example, if you do your trades every day at 3pm, you may want to consider moving them to 1pm, because results show from TCA that you will actually make a lot of money by doing it that way. We are finding people paying a lot closer attention to these types of discussions,” he said.

Is there a danger of TCA becoming a box ticking exercise? “Analysis is only as good as what you use it for,” noted Koutoulas. “You can have it and put it in your DDQ that you’re doing TCA, but if you’re not actually changing anything based on this analysis, there’s not really a value.

“Just like in trading, you use past performance to optimise your models, you change your parameters a little bit so over time you build your sample size,” he continued. “So you can see if you’re consistently underperforming your execution target – it may not necessarily be on price, but on speed – there’s a lot of different metrics within TCA that you can optimise for. I don’t think there’s one overall, perfect execution – and that’s going to differ based on your use case.”

At Ericsson, Samuelsson explained that TCA is used more pre-trade than post-trade. “Pre-trade analysis is really important for us when we build up benchmarks,” he said.

For Wollheim though, TCA can mean different things. “Generally, if you use TCA to just measure the spread and how much you capture of it – that’s one metric – you can use it to measure your different providers,” he said. “We use it a lot in equities to try to measure the different venues we’re on – such as the primary, the dark pool – so it can mean different things and as it evolves over time, it will be really useful as a tool to understand what you’re doing, rather than as a tool for the buy side to show them who to trade with and how.”

What does the future hold for TCA?

While pre-trade TCA seemed to be the way the panellists were making the most use of execution quality analysis, they entertained whether it might one day be used in real time. But again, stretched resources came up. “We try to be as modern as we can with limited resources and limited staff,” said Wollheim. “But we are going to look more at using an algo to do a trade over two hours rather than as a one-off in the market.”

“Is AI likely to play a bigger role in firms’ execution policies,” asked Colin Lambert, managing editor of Profit & Loss. “Are we going to be looking at AI-driven algos that are going to be machine learning on the order so to speak?”

“I think AI is an over-used term,” replied Wollheim, “but if technology gives us the opportunity, sure. I think we’re going to be a lot more automated going forward. We’re not there yet, because we need to spend money in order to automate more and we’re not there yet.”

Turning to Koutoulas, Lambert asked: “Is automation a great thing though? There’s an analogy that you can fly a plane from London to New York without a pilot, but would you get on the first plane without a pilot?”

“You’ve always got to monitor your algorithms – you’re never going to fully remove the human element, because algos are still written by humans. There are different algos for different tasks and I think some brokers and traders are still adding value by knowing which algos to use, how to parameterise them efficiently. We’re not relying on an AI written, black box yet.”

Until that dark day arrives, the panellists generally agreed that when it comes to execution, it’s still important to be able to transact in a number of ways – from traditional voice trading to aggregators to RFQs to algos.

We´d like to use automation in the future as one of many tools of execution. “We’re active in so many markets, in a lot of different time zones, there’s definitely some business we believe are possible to automate, so this is an ongoing discussion,” said Samuelson.

Fear Factor

Lambert raised the question of whether industry standards should be created around algos. “Potentially,” said Wollheim, “I don’t think there’s any magic behind how they work. If you have a limited number of algos that you can use and have a sample of brokers you use them with, pretty soon you’ll see who outperforms and who underperforms – and those are the providers that will see more flow.”

Koutoulas added: “One of my biggest concerns from a market structure standpoint is that, since the rise of e-trading, specialists have gone away, human market makers have gone away, and we haven’t really seen that many high stress events – so we don’t know how algo market makers are going to respond. Their argument is always that they add liquidity to the market, but there’s no regulation or rule requiring that these algos that act as liquidity providers continue to act as liquidity providers in a stress event. So if things start to auto-correlate and flash crash on a prolonged basis, who is to say these LPs are going to be there? I think that’s something that hasn’t been tested in major way.”

As average trade sizes have been slashed in recent years, the type of top of book in FX no longer equates to greater depth of book. Are buy side clients changing how they execute larger tickets now?

“Yes,” said Koutoulas. “There’s the prevalence of icebergs and smaller orders, and orders that aren’t resting as much. You’ve got to be really careful – especially leaving stops. Leaving stops overnight or between open and close when markets are thinner can see huge ranges. Especially if you don’t have defined slippage on your stop. I think that’s a big reason we are seeing these micro flash crashes on a sub-tick level – you’ve got stops in there, liquidity just evaporates there for a tick or half a tick – you’re just to the moon or floor and right back.”

Wollheim suggested that executing larger tickets is an opportunity to work with one’s brokers. “We are in continual discussions with the brokers we use, and they help us understand when we are perhaps trading a ticket that is too large and needs to be divided up. We’ve started working a lot more on leaving orders with brokers and choosing one broker. You need many tools. RFQ is one we use mostly to indirectly prove we haven’t chosen one broker over another, but best execution is just as much about knowing when to choose one broker and knowing when to have an RFQ,” he said.

Samuelsson meanwhile, said his team does a lot more smaller tickets now. “If we did one large ticket before, it now might be a hundred or more, so sometimes we’re splitting it up into lots of small sizes. It was impossible to do book 100 tickets 10 years ago, we because we had to send them to the middle office then through the back office to control and pay everything manually. It was virtually impossible to do that, but now everything flows straight through,” he said.

Samuelsson added that it’s difficult to gauge the market impact of breaking a ticket into 100 tickets. “When it comes to information leakage, we believe that changing how we do it all the time – different products and at different times – can help. We use fixings internally, because it provides us with an easy, visible way of benchmarking ourselves. But if we switch times and periods, that’s the best way not to give away too much. That said, we also understand we can’t keep completely silent,” he said.

Going back to TCA, that’s basically what it’s going to show us, because we can actually dive into the numbers more, added Wollheim. “Like Marcus said, best execution is knowing when to trade, knowing when to spread it out, and knowing when to go to one broker. Even if that’s not the exact wording in the policy, that decision is what best execution is to us,” he said.

Koutoulas added that his firm tries to increase the number of options its portfolio managers have to get their executions done. “For example, we’re doing more on CME Clearport, which is basically off-exchange block trading, and are using some order desks. One thing we’ve noted is that a lot of desks won’t be held to orders anymore, especially if you have a conditional order, so we’ve pulled away a little from them – they don’t want to take the execution risk in fast markets and moved some of that more towards the algorithmic side.”

“It’s early days in TCA in FX, at least for us, so it’s going to be an area where we can improve a lot,” added Wollheim. “We’re seeing some pretty interesting results already. We discuss why a trade may have occurred in a certain way, or why a particular provider gives us that little edge that we want. Even though this isn’t a huge area for us or our portfolio managers, I think it’s extremely important for us to make sure that we don’t lose money for making wrong decisions or for going to the wrong provider.”

Are liquidity consumers asking enough questions of their LPs? “I don’t think we do,” answered Wollheim. “We have quarterly meetings, but could definitely be asking more questions. We just don’t have enough resources, we have to do this in all asset classes and it takes a lot of time.”

“A lot of times, we do post-trade post-mortems: this is what we thought we should’ve had, this is what you gave us. Why? Explain it to us?” said Koutoulas.

Turning to the topic of transparency, Lambert noted that the vast majority of asset managers’ execution costs are in FX – a remarkable fact given that it’s a market in which they don’t often pay attention.

“If you’re coming to market with a large order, whether broken down or not, the more transparent the market is, the more the chance there is for every other market participant to jump in front of it,” suggested Lambert.

“That’s one way of looking at it, but as a general rule the more data and more transparency we have, the better it would be for everyone,” said Wollheim.

Coming back to signalling risk, Lambert asked whether markets are too transparent and if everyone can see the order flow, does everyone come back to the 4pm Fix?  “What happens when the human is removed is that everyone produces a TWAP algo. I can see it happening in the market and I’m not sure that’s best execution,” Lambert noted.

“I would agree, but best execution means different things for the buy and the sell side,” pointed out Wollheim. “For me, it’s the choice of who I trade with and the method. I would look to my brokers and say to them, ‘You have to provide me with the best tools you have, that’s your responsibility’.”

Therefore, Wollheim said, it comes down to how best to maximise performance in the portfolios they run. “That’s why we’re focussing much more on FX than we used to, because there’s an opportunity to enhance profits,” he explained.

Intelligent Execution

Would discussions around intelligent execution be more important then, than best execution? “Yes. For every type of trade, it’s what are you optimising on. There’s no perfect execution, but if you want it fast, if you want to minimise your slippage, if you want to try to outperform the VWAP – it’s about setting the right benchmark and that’s something I think we still need a human for,” noted Koutoulas.

Addressing whether part of the investment manager’s challenge lies in educating the internal surveillance team to understand that best execution must incorporate a variety of nuances. “I feel that our oversight people trust in what we do and trust in the process, and we’re now in a position to be able to verify that with data and I think it has resulted in them having a great deal of confidence in what we’re doing,” said Wollheim, adding that the data is actually contributing to that confidence.

“TCA is in no way perfect,” he continued, “it’s certainly not the answer to everything, but it can provide you with details of things you didn’t know before – patterns and information about things that if you don’t measure it, you have a feeling about what you‘re doing, but can now actually back that up with data.”

But costs can be preventative. According to Samuelsson, “We would love to have that knowledge and possibility and resources to evaluate that, but we don’t. We need to sit down and think about the total business case. We do a lot of business with banks in different areas – corporate finance, M&As, bonds – so somewhere down the line we try to look at the total picture and FX will never be the only important thing for Ericsson, but we need to do it.”


“Are benchmark fixes fit for purpose?” queried Lambert.

“For us, it has been very important in terms of trying to explain what we do and what it means. For the transparency internally, it’s a good product for us. That said, the banks have not been transparent in terms of what it costs,” said Samuelsson.

Wollheim noted that for some of SEB’s managers, primarily in the index space, they want the fixing, because that’s when their portfolio is priced. “If they trade at the fixing, they can just take FX out of the picture completely – even though it may cost them 15 basis points to execute – because they’re in the game to minimise any deviance in the index,” he said, adding that data can help change manager behaviour in this regard.

“If we see a behaviour – in any asset class – that we think isn’t optimal for them, we discuss it with them immediately and, more often than not, they’ll change behaviour,” he added.

Since a lot of managers have an obsession with tracking error, noted Lambert, they may disregard the cost of executing at a benchmark. “Do you think that the benchmark fix is fit for purpose for these guys or should you be part of the conversation?” he asked.

“I’m one of the loudest critics of herd mentality in the industry overall,” replied Koutoulas. “If you look at the concentration of assets at a handful of the largest hedge funds, a lot of behaviour is driven by staying with the herd. A lot of people avoid getting fired that way – if everyone else does it, I’m not the outlier – people like to fire outliers. So anytime you want to change human behaviour, if you combat the norm, you’ve got to show why it’s demonstrably better for them to justify the risk of being the outlier and any costs associated with the time involved in learning a new process or reprogramming or paying a consultant. Big institutions gravitate towards inertia, so it’s hard to change any longstanding behaviour.”

Should benchmarks be more flexible – for example, should they be radically different in the Scandies vs Aussie? “That would definitely be a good thing – that’s something for the industry to work towards,” Samuelsson said.

Lambert asked whether there are some markets in which the benchmark should be your market impact instead of spread? “That’s one of the options when people evaluate TCA – it all comes back to your priorities. For bigger shops, market impact is a big priority,” explained Koutoulas.

“As a general rule, I think measuring market impact is much more important than the actual spread when you execute,” added Wollheim. “The way we’re doing things now, execution doesn’t cost us very much. We have gotten portfolio managers to take much more active positions in the execution stage when they want to do something in one of the other asset classes.”

Last Look

Turning to the rather contentious issue of last look that exists in some circles, none of the panellists voiced concern with the practice.

“Last look is difficult for us to evaluate,” admitted Samuelsson. “As long as our LPs are open about what they’re doing, I guess it’s ok. But since we’re using aggregators a lot, the prices popping up aren’t comparable, which makes it difficult to analyse. But we’re trying to address it through rejected trades and are constantly trying to develop our e- trading feedback.”

Wollheim added: “There’s an impact on TCA to a degree, because the price you’re using for TCA could actually be a price subject to last look, so there’s no guarantee it would be executed anyway. But I don’t have a strong view. It is strange for me given my equity background, but as long as there’s transparency, it’s not huge issue for us. That said, we’d probably use brokers more that didn’t use it.”

“It has been several years since anyone mentioned last look to me – it hasn’t been on my radar in years,” added Koutoulas.

Galen Stops

Share This

Share on facebook
Share on google
Share on twitter
Share on linkedin
Share on reddit

Related Posts in